Adjusting for radiotelemetry error to improve estimates of habitat use.

Abstract

Animal locations estimated from radiotelemetry have traditionally been treated as error-free when analyzed in relation to habitat variables. Location error lowers the power of statistical tests of habitat selection. We describe a method that incorporates the error surrounding point estimates into measures of environmental variables determined from a geographic information system. We estimated a bivariate ellipsoidal probability density for errors surrounding radio collars placed at 20 random sites. This probability density of errors was used to construct probability-weighted estimates of environmental variables. Computer simulations indicated that slope, sine and cosine of aspect, and canopy cover at radiotelemetry locations differed from probability-weighted estimates of those variables (P 0.031). However, these differences were based on large sample sizes (n 305) and were probably too small to influence power of statistical tests of habitat selection. The frequency with which soil, plant community, and canopy cover types were correctly classified with simulated radiotelemetry point estimates increased with increasing patch sizes (P 0.005). Our method could be used to assess how accurately environmental variables can be determined across extremes of habitat and topography and the spatial scale at which analyses retain adequate power. It also could be used with other radiotelemetry systems, including those based on global positioning system technology, if sufficient locations are obtained to describe their probability distribution.

Citation

Findholt, Scott L.; Johnson, Bruce K.; McDonald, Lyman L.; Kern, John W.; Ager, Alan; Stussy, Rosemary J.; Bryant, Larry D. 2002. Adjusting for radiotelemetry error to improve estimates of habitat use. Gen. Tech. Rep. PNW-GTR-555. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station. 14 p